Energy-Efficient Routing Scheme for Distributed Regression in Wireless Sensor Network

This paper presents an energy-efficient routing scheme for the in-network implementation of data regression modelling in wireless sensor networks (WSNs). With the kernel based distributed representation, a clustering based routing architecture is proposed to implement the distributed Gaussian elimination for solving the regression algebraic equation. In particular, the skeleton routing tree is built to coordinate the local computations associated with the clusters. The rest ramose nodes within each cluster are limited to contribute the corresponding local computations. Compared with the junction tree formation over all nodes, the notion of skeleton tree (ST) proposed here suppresses the redundant message passing among the ramose nodes with each cluster without deteriorating data modelling performance. Experimental results are reported to demonstrate the effectiveness of the proposed approach.